If you can access a large enough transaction database, you can map any network to an arbitrary degree of granularity. Having the map, you can plot its topology mathematically. Having the equations & values that describe the topology, you can compute what properties it has - for example, how many nodes need to be cut in order to trigger a cascading failure on a specific subset of the network as a whole. Knowing which nodes need to be manipulated, applying this to real world situations is left as an exercise for the student.
Tim
Tim
posted by:
|
-
Re: maps
Sat, March 11, 2006 - 1:32 AMSorry, this should read:
If you can access a large enough transaction database, you can map any network to an arbitrary degree of granularity. Having the map, you can plot its topology mathematically. Having the equations & values that describe the topology, you can compute what properties it has - for example, how many nodes need to be cut in order to trigger a cascading failure on a specific subset of the network as a whole. Knowing which nodes need to be manipulated in what way to trigger which effects, applying this to real world situations is left as an exercise for the student.
Your answers will be graded on accuracy, thoroughness & inventiveness.
Tim -
-
Re: maps
Sat, March 11, 2006 - 6:10 AMHmmm.
What are you getting at here? Mapping the transactions from an ordinary database application like a supermarket and finding dependencies? Or using a database in general for doing a system analysis?
My guess would be most commercial databases don't have a lot of redundancy and almost any "cut link" or introduced error would cause a cascade of further errors downstream. OTOH, you'd probably prefer a specialist tool built on top of plain Oracle for doing an analysis. (Writing recursive algorithms in SQL sucks!) -
-
Re: maps
Sat, March 11, 2006 - 11:23 AMI'm talking about a database of transactions that describe the traffic going across a network, for instance AT&T's Daytona database, made up of every LD phone call in the US. I'm talking theoretically, not necessarily in practical terms. Implementation & optimization are left as problems for the engineers & programmers to do.
OK, let's step back for a minute. Does anybody have anything that describes a network transform & its effects? Or how about a network topology? Does anybody have anything that takes a bunch of data about a network as input & gives you the topology of the network as output? Let's start there.
Tim
-
-
